Galaxy Zoo Starburst Talk

First go at exploring data: Colors of quench versus control

  • vrooje by vrooje moderator, scientist

    Hi all,

    I've just started working on some color analysis with our sample. Here's my dashboard:
    http://tools.zooniverse.org/#/dashboards/galaxy_zoo_starburst/52011f590aab2a3f3e0000c5

    Some background information

    First: color in an astronomy context is similar to a standard context, but the way you get there from data is to take a ratio of flux (= brightness) from the different SDSS filters, which sample different parts of the (mostly just the visible) electromagnetic spectrum. The filters are, from most blue to most red: u', g', r', i', z'. (Sometimes you can abbreviate and just drop the ' marks.)

    The catalogs show the magnitudes in ugriz, not the fluxes. There's a more detailed description of a magnitude on the SDSS website, but essentially, it's the logarithm of the flux, multiplied by -2.5 (for historical reasons, smaller magnitudes are brighter).

    Because of the rules of logarithms, taking a difference of magnitudes is the same as taking a ratio of fluxes -- so a "color" in astronomy is the difference between magnitudes in two different filters. For example, u-r, or r-z.

    (By convention, astronomers generally subtract the redder filter from the bluer filter; when you do that, higher numbers mean a redder color.)

    Different colors tell you different things about a galaxy, so it's a very useful thing to plot. You can, for example, plot two colors against each other. You can plot a single color versus galaxy mass, or magnitude (preferably "absolute" magnitude, which is intrinsic brightness, so the effects of distance are accounted for; I don't think we have that for these galaxies yet).

    This Dashboard: color-color

    In the dashboard above, I've plotted u-r versus r-z for both the control sample and the quenched sample.

    I had hoped to do more before posting this, but unfortunately I'm about to be buried in a mound of work for the next few days, so I think I should post this now. Hopefully it's a starting point for discussion and for whatever exploration others want to do.

    So, what do you think? See anything interesting, in this or in other colors? (I've made several different colors as new fields in the tables in that dashboard, and I use the convention "m" for "minus", so e.g. "Color_gmr" is the g' - r' color. I just haven't plotted them yet, but you should feel free to!)

    Cheers,

    -Brooke

    Posted

  • mbond by mbond

    Hi,

    With the Histogram tool, one can observe that the distributions of colors (I did it for u-g and r-z) are much narrower for the Quench sample than for the Control one.

    Regards,
    Michel

    Posted

  • klmasters by klmasters scientist

    Thanks Brooke. With some help I copied your dashboard and then decided to make colour-mass diagrams for the two samples. I picked u-r colour as really recently starformation shows up in the u-band, while r-band will reveal more mature stars.

    You can see those two plots here: http://tools.zooniverse.org/#/dashboards/galaxy_zoo_starburst/52021febbe70a6025d000082

    It is nice (although perhaps not surprising) that that reveals that the Control sample shows a red sequence and a blue cloud as should be normal for a sample of galaxies with no other selection, while the quench sample appears to be made up of "green valley" galaxies with medium colours. This suggests to me that they have just recently stopped starforming - as was intended.

    Posted

  • vrooje by vrooje moderator, scientist

    Hi Michel,

    Cool -- what color(s) did you use for the histograms? Karen's idea of u-r is very useful, as that gives an idea of the relative contributions of young and old stars. When we were looking at the color-color diagram, it looked like it was possible the control sample had two populations, which might show up as a histogram with two peaks. Did you see anything like that?

    One thing I did notice that isn't written anywhere on a tutorial is that it's possible to change the name of each tool just by double-clicking on the name. So, for example, I changed the names of my Data tools from Quench-1 and Quench-2 to Quench and Control.

    Posted

  • klmasters by klmasters scientist

    For anyone else going in - I needed telling to minimise windows in the Dashboard to be able to see things. You just hit the little '-' in the top left (and then to reopen the window hit the "o".

    Posted

  • mlpeck by mlpeck

    Here are 2 histograms - first of u-r color, second of d4000. In both histograms the quench sample counts are the red bars and the control counts are the hatched ones. Both (u-r) and d4000 have bimodal distributions in the control sample, while the quench sample distributions are unimodal and much narrower.

    d4000 estimates are from my own "pipeline" btw. They aren't listed in the control sample table.

    enter image description here

    enter image description here

    Posted

  • JeanTate by JeanTate in response to vrooje's comment.

    Dashboard is blank for me vrooje, unlike Karen's

    Posted

  • JeanTate by JeanTate in response to klmasters's comment.

    Very nice Karen!

    There seems to be a hard (u-r) limit, of 3.0, is that because you filtered the data first? Or is it that no galaxy is redder than (u-r) = 3.0?

    Posted

  • JeanTate by JeanTate in response to mlpeck's comment.

    Very nice! 😃

    The first shows the same thing as in Karen's dashboard, only much more cleanly (sans the mass dimension, of course).

    I wonder what the second one shows? That there are essentially two populations (in QC), but QS has galaxies of only one kind (though the QS distribution is narrower than the QC pop1 one, and may also have a somewhat different peak)?

    Posted

  • klmasters by klmasters scientist

    Dn4000 is a measure of the age of the stellar population. It's a break in the spectrum created by lots of metal lines in the atmospheres of low mass stars - you don't see it where there are young stars dominating the light, but it gets revealed as they die off.

    So what we're seeing I think the in the lower histogram is that the Quench sample has Dn4000 like the population of starforming (blue cloud) galaxies in the control sample, not the higher values found in the red sequence. We could try plotting u-r versus Dn4000 for both sample to test that. I'd love to see some morphology folded in too (ie. plot the smooth and featured objects separately, or even hi-light where the mergers and tidal features are). So far this is saying the quench sample are green valley galaxies with young stellar populations (ie. very recently quenched), which is exactly what the aim was when Laura constructed the sample I think.

    Posted

  • JeanTate by JeanTate in response to mlpeck's comment.

    How did you get the d4000 estimates, esp for the QC objects?

    Let's see if I understand this (combined with Karen's post) ...

    AGS00001hn is a QS object, with a d4000 value of 1.36±0.01 in the QS database, which would put it close to the red peak; redshift 0.141.

    AGS00003we is a QC object, with a redshift of 0.155. From its DR9 Explore page, galSpecIndx gives 'd4000' as 1.77±0.04, which would put it near the hatched peak.

    Here are the two spectra, QS object first (QS spectrum, QC); the 4000Ã… break is just red-ward of the the H&K lines.

    enter image description here

    enter image description here

    Hmm, somehow I don't think I've got it right. Not least because the DR9 galSpecIndx value for AGS00001hn is 1.58±0.01! 😮

    Help.

    Posted

  • klmasters by klmasters scientist

    Dn(4000) was defined in a paper by Bruzual (1983) as "the ratio of the average flux density in the wavelength range 4050-4250 and 3750-3950 Angstroms)" A narrow version was defined by Balogh et al. (1999) which uses 3850-3950A and 4000-4100A as the ranges. I don't know for sure as Laura made the catalogues, but the latter has become standard practice, so it's probably that. A good reference for why this is interesting (and where I got the above from - it's not on the tip of my tongue I promise!) is Kauffmann et al. 2003 (Stellar masses and SFH for 10,000 SDSS galaxies).

    Needs to be rest wavelength, not observed wavelength remember.

    Posted

  • mlpeck by mlpeck in response to JeanTate's comment.

    JeanTate:

    As I said elsewhere I have my own data processing "pipeline." Basically I downloaded all the spectra (program and control) and among other things I calculate D4000 using the "narrow version" of Balogh et al. mentioned above. The computation is pretty straightforward and doesn't depend on any sort of SED fitting. Of course my calculations should be used by no-one, including me. What's really needed is to get the same data for the control sample as the program sample, that is emission line fluxes and errors and D4000.

    Posted

  • JeanTate by JeanTate

    Thanks Karen, mlpeck.

    So if we somehow could get D4000 estimates for the QC dataset/catalog, obtained on the same basis as those in the current QS one, and plotted them as you (mlpeck) did, we may see fine differences, but the overall bimodal (QC) vs nice, almost Gaussian distribution (QS) difference would still be there?

    In any case, why are the QS D4000 value and the DR9 galSpecIndx one (for AGS00001hn) so different, not even within 5σ of each other?

    Posted

  • mlpeck by mlpeck

    Yes, I think you'd see the same overall shapes of the distributions. I could give you a detailed statistical comparison of my estimates with the ones tabulated in the QS data table, but it's not really relevant to anything.

    The DR9 galSpecIndx table for AGS00001hn has d4000_n = 1.36 +- .01. That's the "narrow" version mentioned by KL Masters above vs. the original "wide" definition, which I think is what is tabulated as d4000.

    Posted

  • JeanTate by JeanTate in response to mlpeck's comment.

    Thanks.

    I see I made a mistake earlier; I quoted the "d4000" value, not the "d4000_n" one. 😦

    As this SDSS SkyServer page makes clear, d4000 is "Bruzual (1983) definition", while d4000_n is "Balogh et al (1999) definition".

    Posted

  • klmasters by klmasters scientist

    I know it's on Laura T's todo list to upload all the same values for the control sample as the quench sample. So I'm sure Dn4000 will appear very soon for the Control Sample. 😃

    Posted

  • lpspieler by lpspieler moderator

    For Zooites to play around I've created a dashboard with tables for quench sample and control sample that contain all 10 possible colors (differences between channels), pre-filtered Smooth vs. Features:

    http://tools.zooniverse.org/#/dashboards/galaxy_zoo_starburst/52054f83be70a6393f00002b

    As an example of how to play around I added scatter plots for all 4 combinations sample/control and smooth/features of u-r vs. r-z. The difference between quench sample and control sample is remarkable. Even the two "smooth" scatter plots look quite different.

    Posted